DAO - Baidu: Why I Might Be Adding To This AI Leader

Summary

• Baidu was the first to release news on February 7 that it plans to finish its internal testing of its ChatGPT-style project, Ernie Bot, in March.
• Baidu has long been at the forefront of the development of AI technologies in China since its strategic focus on AI in 2016.
• AI technologies will benefit Baidu's AI Cloud and search businesses while allowing third-party applications to benefit.
• Apollo continues to make progress on the autonomous driving front in China.
• My 1-year target price for Baidu is $209, representing 40% upside from current levels. I think that Baidu looks attractive at current levels.

Investment thesis

While Baidu ( BIDU ) has had a tough year in 2022, I think that through 2023, the company's core search business should recover as the Chinese economy strengthens and business activity gains traction with the relaxation of tough Covid-19 measures.

In the long-term, I see Baidu as a core AI business given its strategic focus on AI since 2016. I think that its progress thus far with Ernie and the AI Cloud business has been great, which will help enhance the core search business as well.

Apollo brings a unique EV and autonomous driving opportunity to Baidu, leveraging its strength and expertise in software, AI and technology to leapfrog competition in the space.

I have written an earlier article about AI Cloud and Baidu's EV and autonomous driving opportunity which can be found here .

AI wave in China

Since Google's ( GOOG ) (GOOGL) Bard release on February 6, many other technology companies in China announced that they too are in the process of developing similar AI products and have plans to integrate ChatGPT-like features into their businesses.

Baidu was the first to release news on February 7 that it plans to finish its internal testing of its ChatGPT style project, Ernie Bot, in March. There are reports stating that Baidu plans to make this service a standalone application and these features will be slowly merged into its Baidu search engine, helping generate more relevant search results.

Other companies that announced plans to incorporate ChatGPT features into their businesses include Alibaba ( BABA ), JD.com ( JD ), NetEase ( NTES ) and Tencent ( TCEHY ).

For Alibaba, it states that generative AI and large language models have been its focus since 2017 when it started DAMO . It disclosed that it was in the works of creating a rival to ChatGPT, which is currently undergoing internal testing.

For JD.com, it also plans to launch its own ChatGPT-style product that will be called ChatJD , with focuses on retail and finance and it will have the ability to have human to computer dialogue and generate content.

NetEase also announced that its education business, Youdao ( DAO ), plans to integrate content that is AI generated into its education service.

Lastly, Tencent also stated that research is in progress, and it has plans for ChatGPT-style and AI generated content.

Apart from the companies listed above, I believe that other Chinese technology companies like Huawei and iFlytek are developing similar products.

Some examples of the Chinese technology companies' AI machine learning platform include JD.com's DExplore Dream Team's Vega , Tencent's RoBERTa , Alibaba's PAI , Huawei's NEZHA .

Baidu's advantage in AI technologies

In China, Baidu has been at the forefront of the development of AI technologies, given its many years of experience in the development of technology in the space and the rich expertise they have on AI technologies.

This all began in 2016 when Baidu changed its strategic focus to AI. Here is an excerpt from the CEO in the fourth quarter 2016 earnings call :

In 2016, we began to see glimpses of AI revolutionizing the Internet and traditional industries. The PC era and the rise of mobile over the past four or five years are in our rearview mirror. We are now living and breathing the era of artificial intelligence. We believe that AI is the new electricity, which will transform industry after industry, new applications and products that we have yet to even imagine, and fundamentally, change how users interact with technology. AI represents an enormous opportunity and Baidu is primed to lead the AI revolution in China, as a result of years of investment and our position at the intersection of Big Data, technology and search, as well as our access to the best R&D talent in China and globally.

To seize this opportunity, we have continuously shifted our strategy, organization, and resources to AI. And we continue to gain industry recognition. Launched last September, our artificial intelligence platform, Baidu Brain, is already deployed across nearly all of our products including Search, News Feed, Maps, Nuomi, and PostBar to name but a few.

Since then, with about 6 years of investment focus in AI and deep learning, I think that it is no surprise that Baidu was the first in China to announce an AI chatbot service similar to OpenAI's ChatGPT.

Baidu's Ernie model

Baidu's name for its AI bot is Ernie, which represents "Enhanced Representation through Knowledge Integration." Ernie is a natural language processing ("NLP") deep-learning model that Baidu introduced 2019.

Ernie is a pre-training framework that uses sequential multi-task learning to gradually build and learn through pre-training tasks. Baidu Research announced on 2019 that ERNIE had achieved a new level of performance on GLUE, and it was the first model globally to score over 90 in terms of the macro-average score, with a score of 90.1.

Baidu launched Ernie 3.0 in July 2021, which is the company's unified framework for pre-training large-scale knowledge enhanced models. The model was trained with 10 billion parameters on a 4TB corpus consisting of plain texts and a large-scale knowledge graph.

Ernie 3.0 Titan was then launched in December 2021. It is a pre-training language model with 260 billion parameters and produces good results in both natural language understanding and generation. Baidu thinks that Ernie 3.0 Titan is the first knowledge enhanced multi-hundred billion parameter model globally.

Baidu's AI business opportunity

The first business that Baidu's AI technologies and Ernie are likely to benefit is the search business. Ernie has already been incorporated into Baidu's online products and services such as search, newsfeed, smart speakers, amongst other things.

In Baidu's 1Q19 earning result, management noted that its IME (mobile keyboard) launched an AI version that enables mixed voice input with English, Mandarin and other dialects, real time translation and augmented reality emoticons, and that more than 60% of Baidu IME AI version users opted for voice input.

In addition, the use of AI can improve the search user experience by filtering and eliminating misleading and low-quality advertisements and content. Baidu announced that in the first half of 2019, more than one billion of such content and tens of billions of inappropriate images were filtered.

Secondly, Ernie can also be incorporated into Baidu AI Cloud to improve industrial applications. For example, Baidu improved the predictability in the utility sector using AI as a result of faster visual recognition. Also, in the banking and transport sectors, AI cloud enables more scalable operations and making use of the vast amount of data captured in these businesses. Ernie currently covers 11 industries, including finance, utilities and manufacturing, amongst others.

Lastly, Baidu could release their AI technology to support third party applications, whether it involves gaming, designers or coding.

Baidu Core

In terms of Baidu App user metrics, despite its large user base, the MAU grew 8% year on year to 646 million and DAY grew 5% year on year to 240 million, as of December 2022 based on data from Questmobile. As of December 2022, total time spent was up 8% year on year to 319 billion minutes, while the average user time spent reached 494 minutes, which is flat year on year.

Based on a Reuters report, it was stated that critically ill Covid-19 cases in China was down 72% , from the peak of 128,000 cases on Jan 4 to 36,000 cases on Jan 23. It was also stated that over 80% of China's 1.4 billion population had already been infected.

With the drastic change in the policy for Covid-19 in China, I think we will see that online advertisements were likely impacted by slower economic activities as a result of the earlier tighter measures, which has been largely priced in, in my view.

In terms of 2023's outlook, I expect that the first quarter could see a slower start, but business momentum should build, especially after the Lunar New Year.

Baidu Cloud

Baidu held its annual Create AI Developer Conference on January 10th to showcase its latest developments in AI and cloud technology. Mr. Robin Li, Baidu's CEO and chairman reiterated that the company is one of the few companies that have invested heavily into AI and the cloud. Baidu's AI technology framework is expected to target and lower the technological barrier for the adoption of AI.

Baidu's unique integration of AI and cloud enables more companies to access AI technology, which helps to improve operating efficiency for customers. Baidu continues to be recognized as a leader in AI public cloud, conversational AI and AI industrial quality control verticals, based on third party research.

Baidu made two major moves in 2022 in its AI cloud business. The first was the introduction of integration of cloud and AI 3.0 frameworks and the in-house developed AI infrastructure AI Stack. Baidu AI Stack represents Baidu Cloud's AI IaaS, which integrates its Kunlun chips and optimizes the performance in AI compute, storage, and acceleration. On AI PaaS layer, combining PaddlePaddle deep learning framework and Wen Xin model, Baidu also offers full-chain solutions for corporate model development.

All in all, Baidu is well positioned to help companies lower the barrier to entry into adopting AI.

Baidu's autonomous driving

During the Baidu Create Conference on January 12, Apollo unveiled the new BEV autonomous driving sensation solution. This solution aims to improve the detection capabilities of intelligent vehicles by using advanced sensors like RIDAR and HD devices, which provide a better perception of the surrounding environment than human vision. The data collected by these sensors will be fed into an integrated model to generate more accurate results. The BEV solution also supports time series integration to analyze data over time, resulting in more stable and reliable results.

Apollo also introduced the UniBEV solution, which integrates multiple sensors, models, and missions to provide end-to-end autonomous driving solutions. Baidu also unveiled its smart road OS, which features five key functions, including AI middle office, navigation services, communication, open-source application service capability, and hardware connection.

In April 2021, Baidu launched the Apollo AIR autonomous driving system for data integration of vehicles and the road environment, which supports open road sensation at L4 autonomous driving level. This technology has also been applied in urban assistance driving, smart information control, and valet parking in over 60 cities in China by the end of 2022.

Valuation

Using a sum of the parts model, I derive my 1-year target price for Baidu given that the company has several different business segments growing at different rates. My assumption for Baidu AI cloud business is a 4x multiple on 2023 sales and 8x 2023 P/E for the core search business, while I use a DCF model for the Apollo and intelligent vehicle business. Lastly, I use the fair value of its investments based on their most recent valuation and apply a 30% holding company discount.

My 1-year target price for Baidu is $209, representing 40% upside from current levels. I think that Baidu looks attractive at current levels.

Risks

Weak macro backdrop

I think that despite China's re-opening and relaxing of restrictions, there is the risk that if the global macro backdrop weakens, this could affect business and consumer sentiment in China. In particular, its search business will be affected given the lower advertising spend in such a scenario, providing near-term headwinds to the business.

Delay in progress of launch of EV and autonomous driving milestones

Baidu continues to execute on its autonomous driving plans. That said there are risks that there might be delay in the launch or the company may have problems ramping production up. There is a need to make sure that the company executes well on its autonomous driving business timeline given the increasing relevance for Apollo in Baidu's valuation.

Conclusion

All in all, I think that Baidu should see demand for ads recover through 2023 as the economic activity in China kicks off after a long period of restrictions. This will be positive for its core search business in the near-term.

In the long-term, Baidu is focused strategically on investing in AI and its progress that it has been making in its AI Cloud business and Ernie is encouraging, which will have long-term benefits to its core search business as well. Lastly, Apollo presents a unique EV and autonomous driving opportunity to a company traditionally known more for search. My 1-year target price for Baidu is $209, representing 40% upside from current levels.

For further details see: